Katharina J. Rohlfing
2014
A multimodal corpus for the evaluation of computational models for (grounded) language acquisition
Judith Gaspers | Maximilian Panzner | Andre Lemme | Philipp Cimiano | Katharina J. Rohlfing | Sebastian Wrede
Proceedings of the 5th Workshop on Cognitive Aspects of Computational Language Learning (CogACLL)
Judith Gaspers | Maximilian Panzner | Andre Lemme | Philipp Cimiano | Katharina J. Rohlfing | Sebastian Wrede
Proceedings of the 5th Workshop on Cognitive Aspects of Computational Language Learning (CogACLL)
2011
Exploring "Associative Talk": When German Mothers Instruct Their Two Year Olds about Spatial Tasks
Katharina J. Rohlfing
Dialogue Discourse Volume 2
Katharina J. Rohlfing
Dialogue Discourse Volume 2
In this study, maternal input was analyzed during a task, in which German mothers instructed their two-year-old children to put two objects together in a particular way. In the setting, the spatial relation (ON and UNDER) and the canonicality of these relations (canonical such as ‘a pot on a table’ and noncanonical like ‘a train on a tunnel’) were varied. Two kinds of discourse strategies are proposed that characterize mothers’ input in this task: bring-in and follow-in. For the analysis, an automatic procedure was developed, in which the amount of words spent on a strategy was related to the overall word amount. The data suggest that the canonicality of the task can change the discourse: Bring-in strategies dominated the discourse in tasks with canonical spatial relations while in more difficult tasks with non-canonical relations, German-speaking mothers used follow-ins significantly more often than in the canonical tasks. Together, the results of this study shed light on the process of an on-line adaptation of the mother to her child and give us insight into how a situated understanding in a task-oriented discourse emerges.